Course 2.4 Perception 1
Problem 5: Constituent Parts and Their Wholes
Literature: Wolfe, Goldstein (psyweb), Gazzaniga
How do we recognize?
- First: basic visual features (Simple cells/complex cells/end-stopped cells)
- Then: combine information
→ no cell sees whole picture
Middle Vision:
- Combines features to objects
- Goal: organize elements of a scene into groups → then recognize objects
Finding edges
- Still able to find edge even though there is a “gap” (harder to find the gap)
o Computer: finds gap
o Visual system: knows “gap” is due to lighting → fills in missing info
- Scenes with much detail → visual system finds meaningful edges; contour
o Computer: finds all edges; fine details
- Contour perception is of inferential nature (e.g. Kanizsa figure)
o Easy to see arrow outline even though it´s not explicitly drawn
o illusory contour = best guess to make sense of what is happening
- Rules of evidence
o Structuralist: “Perception = sum of atoms of sensation” (→does not allow illusory contour to happen)
o Gestalt: “Perception is not only sum of sensory information. Organizing principles describe visual
interpretation of raw retinal images.”
= Gestalt grouping rules (reflect regularities in the world)
E.g. for rule: similarly oriented lines are grouped → lines
“support” each other because they are on 1 contour
When lines form closed shape → grouping seems
stronger
Good continuation = grouped together if they seem to lie on
the same contour
- Perceptual “committees”: Metaphor for complex, intertwined network of neurons
o “all else being equal” – all guesses of perceived shape being equal (?)
o “Committee” model: all parts come together and give opinion on how stimulus should be understood. →
come to consensus; settle on 1 interpretation
o Middle vision behaves like many specialists (own views; oppinions…)
Must agree on one answer
o Selfridge (1959): “Pandemonium” model for letter recognition
Committee members = “demons”
“feature demons” find features (e.g. vertical/horizontal/curved lines; angles)
“cognitive demos” have ideas about features of letters. Makes noise proportional to similarity of
feature.
“decision demon” listens to the committee, identifies the loudest yell
o Rules: Honor Physics, avoid accidents:
Ambiguous figure = has more than 1 way of interpretation (e.g. Neckar Cube/ Duck-Rabbit)
Technically every figure is ambiguous. Usually committee comes to an agreement
Accidental viewpoints: one viewing position that leads to a regular/lined up image
Chaning viewpoint → destroys illusion
Visual system knows about these → doesn’t often assume that’s happening
o Separate committees for opaque objects, accidental viewpoints
- Occlusion: something may be covering the contour
Problem 5: Constituent Parts and Their Wholes
Literature: Wolfe, Goldstein (psyweb), Gazzaniga
How do we recognize?
- First: basic visual features (Simple cells/complex cells/end-stopped cells)
- Then: combine information
→ no cell sees whole picture
Middle Vision:
- Combines features to objects
- Goal: organize elements of a scene into groups → then recognize objects
Finding edges
- Still able to find edge even though there is a “gap” (harder to find the gap)
o Computer: finds gap
o Visual system: knows “gap” is due to lighting → fills in missing info
- Scenes with much detail → visual system finds meaningful edges; contour
o Computer: finds all edges; fine details
- Contour perception is of inferential nature (e.g. Kanizsa figure)
o Easy to see arrow outline even though it´s not explicitly drawn
o illusory contour = best guess to make sense of what is happening
- Rules of evidence
o Structuralist: “Perception = sum of atoms of sensation” (→does not allow illusory contour to happen)
o Gestalt: “Perception is not only sum of sensory information. Organizing principles describe visual
interpretation of raw retinal images.”
= Gestalt grouping rules (reflect regularities in the world)
E.g. for rule: similarly oriented lines are grouped → lines
“support” each other because they are on 1 contour
When lines form closed shape → grouping seems
stronger
Good continuation = grouped together if they seem to lie on
the same contour
- Perceptual “committees”: Metaphor for complex, intertwined network of neurons
o “all else being equal” – all guesses of perceived shape being equal (?)
o “Committee” model: all parts come together and give opinion on how stimulus should be understood. →
come to consensus; settle on 1 interpretation
o Middle vision behaves like many specialists (own views; oppinions…)
Must agree on one answer
o Selfridge (1959): “Pandemonium” model for letter recognition
Committee members = “demons”
“feature demons” find features (e.g. vertical/horizontal/curved lines; angles)
“cognitive demos” have ideas about features of letters. Makes noise proportional to similarity of
feature.
“decision demon” listens to the committee, identifies the loudest yell
o Rules: Honor Physics, avoid accidents:
Ambiguous figure = has more than 1 way of interpretation (e.g. Neckar Cube/ Duck-Rabbit)
Technically every figure is ambiguous. Usually committee comes to an agreement
Accidental viewpoints: one viewing position that leads to a regular/lined up image
Chaning viewpoint → destroys illusion
Visual system knows about these → doesn’t often assume that’s happening
o Separate committees for opaque objects, accidental viewpoints
- Occlusion: something may be covering the contour
